CSM6120 Introduction to Intelligent Systems Introduction to the module [email protected] Commitment 20hrs seminars 6hrs practical Rest of time spent background reading and on assignments/presentation 4hrs seminars allocated to group presentation prep Assessment: Assignment 1= 40% Assignment 2 = 60% (presentation + report) (coding + report) How to succeed in this course Manage your time wisely Previous students have sometimes struggled with the intensive nature of the course Plan for the deadlines Do lots of background reading, and read around the subject If you’re stuck, then ask! Ask a fellow student or ask me Module content Introduction to AI Search – uninformed and informed Knowledge representation Propositional and First-Order Logic Rule-based systems Knowledge acquisition Neural nets and subsymbolic learning 1. 2. 3. 4. 5. 6. 7. Course notes etc will be made available in: http://www.aber.ac.uk/~dcswww/Dept/Teaching/CourseNotes/20122013/CSM6120/ Timing Today: CSM6120 starts This week: Assignment 1 handed out Next week: Assignment 2 handed out October 19th CSM6120 teaching ends Presentations on Thursday 18th Assignment 1 due in on the Friday 26th November 2nd CSM6120 assignment 2 deadline Timing 1 prep 2 prep pres Hand in 1 Hand in 2 Book list Russell, S. and Norvig, P. - Artificial Intelligence : a modern approach, 3rd edn, Prentice Hall, 2010 (previous editions just as useful, though there have been a few amendments) (first chapter: http://www.eecs.berkeley.edu/~russell/intro.html) Luger, G. - Artificial intelligence : structures and strategies for complex problem solving, Pearson Addison-Wesley, 2009 Coppin, B. - Artificial Intelligence Illuminated, Jones and Bartlett Publishers, 2004 etc... What is Artificial Intelligence? Understand intelligent entities Build intelligent entities Learn more about ourselves/animals Create things that exhibit ‘intelligence’ Study constructed intelligent entities These constructed entities are interesting and useful in their own right! What is Artificial Intelligence? Scientific Goal To determine which ideas about knowledge representation, learning, rule systems, search, and so on, explain various sorts of real intelligence Engineering Goal To solve real world problems using AI techniques such as knowledge representation, learning, rule systems, search, and so on AI problems Formal tasks - playing board or card games, solving puzzles, mathematical and logic problems Expert tasks - medical diagnosis, engineering, scheduling, computer hardware design Mundane tasks - everyday speech, written language, perception, walking, handling What is Artificial Intelligence? “Artificial Intelligence (AI) is the part of CS concerned with designing intelligent computer systems, that is, systems that exhibit characteristics we associate with intelligence in human behaviour – understanding language, learning, reasoning, solving problems, and so on.” (Barr & Feigenbaum, 1981) “The study of the computations that make it possible to perceive, reason, and act” (Winston, 1992) “The branch of computer science that is concerned with the automation of intelligent behaviour” (Luger and Stubblefield, 1993) History of AI 1943: Warren Mc Culloch and Walter Pitts: a model of artificial boolean neurons to perform computations First steps toward connectionist computation and learning (Hebbian learning) Marvin Minsky and Dean Edmonds (1951) constructed the first neural network computer Made out of 3000 vacuum tubes and a surplus automatic pilot mechanism from a B-24 bomber Simulated a network of 40 neurons 1950: Alan Turing’s Computing Machinery and Intelligence First complete vision of AI Anticipated all major arguments against AI in following 50 years History of AI 1956: Dartmouth Workshop Brings together top minds on automata theory, neural nets and the study of intelligence Allen Newell and Herbert Simon: the logic theorist (first non-numerical thinking program used for theorem proving) Proved 38 of the first 52 theorems in Principia Mathematica, found more elegant proofs for some For the next 20 years the field was dominated by these participants 1952-1969 Newell and Simon introduced the General Problem Solver: imitation of human problem-solving Arthur Samuel investigated game playing (checkers) with great success John McCarthy (inventor of Lisp) Logic oriented, Advice Taker (separation between knowledge and reasoning) History of AI The first generation of AI researchers made these predictions about their work: 1957, Simon and Newell: "within ten years a digital computer will be the world's chess champion" and "within ten years a digital computer will discover and prove an important new mathematical theorem." 1965, Simon: "machines will be capable, within twenty years, of doing any work a man can do." 1967, Marvin Minsky: "Within a generation ... the problem of creating 'artificial intelligence' will substantially be solved." 1970, Marvin Minsky: "In from three to eight years we will have a machine with the general intelligence of an average human being.“ Expectations were high! History of AI Collapse in AI research (1966 - 1973) Progress was slower than expected Some systems lacked scalability Unrealistic predictions Combinatorial explosion in search Fundamental limitations on techniques and representations Minsky and Papert (1969) Perceptrons Research funding for neural net research soon dwindled to almost nothing History of AI AI revival through knowledge-based systems (1969-1970) General-purpose vs. domain specific Expert systems MYCIN to diagnose blood infections (Feigenbaum et al.) – introduction of uncertainty in reasoning Increase in knowledge representation research E.g. the DENDRAL project (Buchanan et al. 1969) – first successful knowledge intensive system (large numbers of rules) Logic, frames, semantic nets, … AI winter (1974-1980) Lighthill report highly critical of some areas History of AI AI becomes an industry (1980 - present) Connectionist revival (1986 - present) XCON at DEC (1980) – saved the company $40m p.a. Fifth Generation Project in Japan (1981) – $850m to build machines that could make conversations, translate languages, interpret pictures, and reason like humans Parallel distributed processing (Rumelhart and McClelland,1986); backpropagation Symbolic models vs connectionism AI becomes a science (1987 - present) History of AI 1990s Emergence of intelligent agents: bots! Machine learning Genetic algorithms 2000+ Dealing with large datasets Swarm intelligence ... Large field, lots of applications AI and Games Classic Games Noughts and Crosses Chess - Deep Blue 1997 1957 - Newell and Simon predicted that a computer would be chess champion within ten years Simon : “I was a little far-sighted with chess, but there was no way to do it with machines that were as slow as the ones way back then” Connect 4, Othello, Backgammon, Scrabble, Bridge, Go Current Games Strategy/Tactical/Combat (F.E.A.R., Crysis) RPG/Adventure Artificial Life (Creatures, Spore) AI approaches Thinking vs Acting Human vs Rational (acting = behaviour) (rationality = doing the right thing) Systems that think like humans Systems that think rationally (cognitive science) (logic/laws of thought) Systems that act like humans Systems that act rationally (c.f. Turing test) (agents) Artificial Intelligence AI often burdened with over-promising and grandiosity The gap between AI engineering and AI as a model of intelligence is so large that trying to bridge it almost inevitably leads to assertions that later prove embarrassing McCarthy said AI was “the science and engineering of making intelligent machines” So how can we determine if a program is intelligent? Strong vs Weak AI Debate as to whether some forms of AI can truly reason and solve problems Strong AI: Machine can actually think intelligently Weak AI: Machine can possibly act intelligently John Searle “...according to strong AI, the computer is not merely a tool in the study of the mind; rather, the appropriately programmed computer really is a mind” Turing Test (1950) Human interrogator Human ? AI System Turing's argument is essentially: “If a computer can fool a judge into thinking it is human, we must acknowledge it is able to think like a human” Turing Test (1950) What techniques are required? Natural language processing to enable it to communicate successfully in English (or some other human language) Knowledge representation to store information provided before or during the interrogation Automated reasoning to use the stored information to answer questions and to draw new conclusions Machine learning to adapt to new circumstances and to detect and extrapolate patterns Turing Test (1950) AI researchers have devoted little effort to passing the Turing test Believe that studying principles of intelligence is more important than duplicating something else Precedent? The quest for artificial flight Succeeded when people stopped imitating birds and learned aerodynamics Aeronautical engineering does not define its goal as making “machines that fly so exactly like pigeons that they can fool even other pigeons” Chinese Room Searle argued that behaving intelligently was not enough (1980) Thought experiment - the Chinese Room You are in a room with an opening through which Chinese sentences are passed You have a rule book that allows you to look up these sentences although you do not speak Chinese The book tells you how to reply to them in Chinese You can then behave in an apparently intelligent way (video) Chinese Room Searle claimed that although they appeared intelligent, computers would be using the equivalent of a rule book The rule book and stacks of paper, just being paper, do not understand Chinese Within the article setting out the Chinese Room experiment, Searle set out some possible arguments against his contention that the individual in the Chinese Room could not be said to understand What does it all mean? The Chinese Room argument has provoked much discussion Watson In 2011, Watson beat the two most successful Jeopardy players http://www.bbc.co.uk/news/technology-12491688 http://www.bbc.co.uk/news/technology-17547694 But is this intelligence??? DeepQA article: http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=617 7810 http://www.aaai.org/Magazine/Watson/watson.php Ethics and AI We’ve looked at whether we can develop AI, but not whether we should The problems that AI poses: People might lose jobs to automation People might have too much/little leisure time People might lose some of their privacy rights Loss of accountability – who’s to blame if things go wrong? Success of AI might mean end of human race! Almost any technology has the potential to cause harm in the wrong hands Branches of AI (John McCarthy) Logical AI: What a program knows about the world in general the facts of the specific Search: AI programs often examine large numbers of possibilities, e.g. moves in a chess Pattern recognition: When a program makes observations of some kind, it is Representation: Facts about the world have to be represented in some way. situation in which it must act, and its goals are all represented by sentences of some mathematical logical language. The program decides what to do by inferring that certain actions are appropriate for achieving its goals. game or inferences by a theorem proving program. Discoveries are continually made about how to do this more efficiently in various domains. often programmed to compare what it sees with a pattern. For example, a vision program may try to match a pattern of eyes and a nose in a scene in order to find a face. More complex patterns, e.g. in a natural language text, in a chess position, or in the history of some event are also studied. These more complex patterns require quite different methods than do the simple patterns that have been studied the most. Usually languages of mathematical logic are used. Branches of AI (John McCarthy) Inference: From some facts, others can be inferred. Mathematical logical deduction is Commonsense knowledge and reasoning: This is the area in Learning from experience: Programs do that. The approaches to AI based adequate for some purposes, but new methods of non-monotonic inference have been added to logic since the 1970s. The simplest kind of non-monotonic reasoning is default reasoning in which a conclusion is to be inferred by default, but the conclusion can be withdrawn if there is evidence to the contrary. Ordinary logical reasoning is monotonic in that the set of conclusions that can the drawn from a set of premises is a monotonic increasing function of the premises. which AI is farthest from human-level, in spite of the fact that it has been an active research area since the 1950s. While there has been considerable progress, e.g. in developing systems of non-monotonic reasoning and theories of action, yet more new ideas are needed. on connectionism and neural nets specialize in that. There is also learning of laws expressed in logic. Programs can only learn what facts or behaviours their formalisms can represent, and unfortunately learning systems are almost all based on very limited abilities to represent information. Branches of AI (John McCarthy) Planning: Planning programs start with general facts about the world (especially facts Epistemology: This is a study of the kinds of knowledge that are required for solving Ontology: Ontology is the study of the kinds of things that exist. In AI, the programs Heuristics: A heuristic is a way of trying to discover something or an idea embedded in Genetic programming: Genetic programming is a technique for getting about the effects of actions), facts about the particular situation and a statement of a goal. From these, they generate a strategy for achieving the goal. In the most common cases, the strategy is just a sequence of actions. problems in the world. and sentences deal with various kinds of objects, and we study what these kinds are and what their basic properties are. Emphasis on ontology began in the 1990s. a program. The term is used variously in AI. Heuristic functions are used in some approaches to search to measure how far a node in a search tree seems to be from a goal. Heuristic predicates that compare two nodes in a search tree to see if one is better than the other, i.e. constitutes an advance toward the goal, may be more useful. programs to solve a task by mating random programs and selecting fittest in millions of generations.